●builderYou can integrate this into your agentic workflows to catch misspecified tasks that would otherwise cause agents to manipulate test environments.
●policyThis provides a verifiable mechanism to ensure autonomous agents operate within formally defined safety constraints.